Working Life of Infectious Disease Models - Dr Erika Mansnerus
Working life ofinfectious disease models Dr Erika Mansnerus London School of Economics and Political Science
Working life of models• Processes and practices when models are in use; when they ‘make things happen’• Working at a distance… – Evidence eventually informs policy, can be contested (e.g. Oreskes and Conway 2010 on Climate research)• …Or very close – Results taken into account prior to final publication (case on measles modelling in the UK 1990s)• Discursive space (Evans 2000) – Dialogue with different actors in order to create shared understanding
Why measles booster campaign?• 1988: Introduction of the MMR triple vaccine• Incidence of measles in England and Wales declined to all time low• Characteristic seasonal and 2-yearly cycles have disappeared• But, measles persists – Should the single dose campaign be enough or is a policy change justified?
1994 National measles and rubella campaign in the UK• Why? - Meant to prevent an epidemic by vaccinating 95% of 7 million schoolchildren (5-16 year olds)• Cost-effectiveness of a preventive action• Based on epidemiological surveillance data – Serological studies – Number of notified and confirmed cases – Rates of complications and deaths – Immunisation coverage• Data used in two independent mathematical models that predict high probability of a major “resurgence of measles” – Resurgence result of poor vaccine coverage and vaccine failure (Miller 1994)
Responding to policy• Two independent models were built prior to the introduction of the 1994 campaign – 1) Gay, N., Hesketh, L., Morgan-Capner P., and E. Miller (1995): WAIFW-model (who acquires infection from whom) – 2) Babad, H., Nokes, D., Gay, N., Miller, E, Morgan- Capner P., and R.M. Anderson (1995): RAS-model (a realistic age-structured measles transmission model)
Interpreting data with WAIF-model• Gay et al (1995): Models used to interpret susceptibility data from the surveillance programme and evaluate the potential for an epidemic.• The population is divided into several age-groups and the transmission rates between these groups are derived from pre-vaccination case notification data.• The transmission rates are combined with data susceptibility in the population to generate a ‘next generation matrix’ = WAIFW matrix (contains values of the transmission rates between groups)• Two different scenarios are modelled with different reproduction rates
Outcomes from the WAIFW-model• Increasing population of schoolchildren susceptible to measles provided the potential for a major epidemic of measles in the mid 1990s.• This prediction was confirmed by using other data (disease notifications from England and Wales 1993/4)• The vaccination campaign is expected to have a “dramatic effect” on susceptibility to measles• New role for models in public health setting: applying them to interpret serological surveillance data allows the potential for an epidemic to be identified, and provides time to plan and implement appropriate interventions
The RAS-model• A realistic, age-structured mathematical model of measles transmission to reconstruct the impact of measles vaccination in England and Wales from 1968 to the present and to evaluate the merits of future policy options.• Good agreement between the observation and prediction, hence model is used to explore measles epidemiology in the future under existing vaccination strategy and a variety of alterations• Notification data is ‘messy’: age-specific bias in reporting; the proportion of misdiagnosed cases.
Outcomes from the RAS-model• Predictive capacity of a mathematical model of the impact of mass vaccination on the epidemiology of infectious disease• It has not been possible to fully account for the discrepancies between the observed impact of vaccination on measles and that predicted• Model projections suggest that the current policy of immunization may not be sufficient in to eliminate measles• One-off campaign targeted at school-age children would reduce deficits in the herd immunity profile, depress the potential for seasonal outbreaks, thus enhancing the effect of current policy
Discussion: Benefits and limitations• Able to interpret serological data and predict outbreak – Planning interventions before the outbreak – Integrating various sources of evidence in the process• Modelling along with decision-making feeds results and evidence directly to the process – Advice through modelling• Not able to address vaccine compliance• Discrepancy between observed and predicted impact of vaccinations
Acknowledgements• British Academy Post-Doctoral Fellowship held jointly at CRASSH, and LSE Health (2009-2012)• Health Protection Agency, Centre for Infections, Colindale• The National Institute for Health and Welfare, Helsinki